HackOn(Data) 201622 Jan 2017
For our third-place project description, click here
For the video of our presentation, click here
Last September, I participated in HackOn(Data), a two-day data hackathon in Toronto. It was the first year it was held and one of the few data science competitions in Toronto. I learned a lot, met similar-minded data enthusiasts and even ended up winning third-place with my teammate Chris!
In the two months leading up to HackOn(Data), I completed weekly online and in-person workshops that built a prerequisite knowledge of Apache Spark (on Databricks), machine learning and general data handling. I really enjoyed these workshops as they guided us through practical problems and techniques involving real datasets. Participants also competed against each other for points, which could be attained through completion of workshops and challenges, and other promotional activities. The top 5 participants received prizes at the HackOn(Data) event, but unfortunately, I finished in seventh place, shy of the prizes.
During the last in-person workshop before HackOn(Data), I met Chris Goldsworthy, a fellow University of Toronto Student in CS, and we decided to form a team. The challenges and datasets for the main HackOn(Data) event were released a week in advance, so Chris and I met up in the days before the event to discuss details. We chose to do the challenge that we thought was the most interesting and open-ended: an unsupervised learning project to determine optimal map placements in Toronto based on datasets of cultural centre locations, transit stops and traffic.
We then brainstormed ideas and created preliminary models for our challenge, which was crucial for us as we quickly eliminated bad or infeasible ideas and set up a concrete workflow. Had we waited until the event, it would have been much more difficult to find our path and complete meaningful work, due to the unsupervised nature of our challenge and the slim 24 hours we had to work on our project during the event. It was at this point that we came up with the idea to augment the existing datasets with social media data mined from Foursquare, a unique feature of our project.
Saturday morning, I arrived at Wattpad HQ, the event location of HackOn(Data), by foot and late, regretting my decision to walk 3km in new dress shoes. After formalities, presentations and a nutritious lunch, Chris and I locked down a secluded place to work in the upstairs loft, furnished with sofas and a spare mattress. We met with our project mentor, Ryan, a City of Toronto geospatialist expert, to discuss ideas and receive useful feedback. At HackOn(Data), there were an abundance of mentors; at times, it seemed like there were more mentors than participants, which meant that there was always plenty of help available.
Chris and I spent the remainder of the day hacking and snacking until it was time to go home for the night. Unlike traditional hackathons, geared towards sleepless university students, HackOn(Data) was meant more for professionals, and hence closed down for the night, to resume in the morning.
The deadline for the submission was the next morning so Chris and I continued our work back on UofT main campus that night and in the early morning to finish our project and prepare our presentation. We arrived back at Wattpad HQ Sunday morning exhausted but finished and content with our work. After breakfast and more sponsor talks, it was time for the first round of judging.
During the initial round of presentations, Chris and I gave a very quick overview of our method and code to a panel of 6 judges and did well enough to progress to the final four round. We had roughly an hour to prepare our final pitch, which was to be given to everyone at HackOn(Data), including a panel of 4 finalist judges.
We presented third. Our presentation went smoothly but our question period was long and tough, and we received many questions on the methods we used that went beyond our original task, such as mining social media data from Foursquare and a scoring system to rank our final placements (see project page). After some intense grilling, we concluded our presentation and felt a great surge of relief knowing that the most stressful part was over.
After some deliberation by the judges, the results were announced: We won third place! Chris and I were very happy about our result, considering that we were just two undergraduate students, relatively new to data science, competing against industry professionals and PhDs. For our efforts, we received $300 in cash, two Amazon Kindles, and various small gifts.
With the awards ceremony, the HackOn(Data) event was over, but the fun was just about to begin. After closing remarks, we went to have drinks and food at a nearby restaurant, courtesy of HackOn(Data). Perhaps the most important part of HackOn(Data), it was here where we chatted with bigwigs from Flipp, Wattpad, IBM and the City of Toronto, as well as graduate students with similar desires to enter the data science industry. We received useful advice about the data science industry, and made important connections that strengthened our professional networks. The amazing part is that Chris and I both received summer internships with HackOn(Data) sponsor companies, whom we met at the event! With third place prizes in our hands and smiles on our faces, we departed the event, ready to begin the school semester the next day.
An exceptional event, well-organized by Mehrdad and his crew, with numerous sponsors and mentors. Would have liked the possibility to work overnight at Wattpad HQ, but understandably, the overhead cost is huge.
In-person and online workshops in the weeks leading up to HackOn(Data) taught me a great deal about Apache Spark and general ML techniques. The points system was an excellent motivator for me to complete the weekly workshops.
Great prizes but perhaps the best rewards were the experience we gained and the connections we made to industry professionals.
Good food, comfortable spaces, great people.
See you at HackOn(Data) 2017!